arXiv:0905.0074v1 [quant-ph] 1 May 2009 igepoo ttswierjcighge htnnum- only higher 17]. rejecting transmit [16, while that effect states re- photon filters Zeno measurement single photonic the the sophisticated for as More as such out well effects read as and lated initialization the photonic Polariza- for 1A). of crucial horizontal are (Fig. the filters only tion incident transmits the that of component a is qubits communication. quantum and relays distance quantum long cre- realizing for 15], the in repeaters [14, for important entanglement be of used purification will be the which can as filter well as The informa- ation to [13]. and approach leading processing 11] a tion [6, are metrology and logical 10], [12], the lithography [1, are communication photons for as promis- choice technologies, is quantum it for qubits, our photonic ing As filter on the acts ones. filter of classical entanglement from capability beam it entangling distinguishing The polarizing verified, was 9]. partially 8, and [7, displaced- splitters [6] approaches—a architecture technological Sagnac recent key two lacking. been have several interferomters, in nested interference ancil- classical both two and requiring with interference gates, quantum circuit quantum optical multiple and an for photons device, lary a proposed requirements such been technical build the has to however, device [5], a qubits photonic Such typically qubits. systems, two-level quantum multiple op- on must quan- non-locally entanglement erate the with on associated mechanical act correlations which quantum tum Filters these entanglement. of is [4]. unique features lithography most and the [1], [3], Perhaps new communication measurement and in [2], improvement tasks computation exponential particular for gain mechan- functionality quantum to harnessing effects tech- with ical and concerned science is tools information Quantum nology technological and one us. scientific to are important available etc) most the , of frequency, signal, (material, 3 h otcmo xmlso le o photonic for filter a of examples common most The combining by filter entanglement an demonstrate We unwanted the reject and desired the pass that Filters, etefrQatmPoois .H il hsc Laborator Physics Wills H. H. Photonics, Quantum for Centre ASnumbers: appli i PACS important photons many of have pair devices a Such passes polarization. that their filter entanglement optical w wit filters an found multi- have to extension , requires or entanglement filters, one-qubit where nology, h blt ofitrqatmsae sakycpblt nqua in capability key a is states quantum filter to ability The nvriyo rso,Mrhn etrr ulig Woodl Building, Venturers Merchant Bristol, of University 1 eerhIsiuefrEetoi cec,Hkad Unive Hokkaido Science, Electronic for Institute Research y Okamoto, Ryo ooiaNagata, Tomohisa sk nvriy ioak -,Iaai sk 567-0047, Osaka Ibaraki, 8-1, Mihogaoka University, Osaka 2 4 h nttt fSinicadIdsra Research, Industrial and Scientific of Institute The iohm nvriy iohm 3-50 Japan 739-8530, Hiroshima University, Hiroshima rdaeSho fAvne cecso Matter, of Sciences Advanced of School Graduate ,2, 1, nEtnlmn Filter Entanglement An ∗ ,2 1, eeyL O’Brien, L. Jeremy ej Sasaki, Keiji etpoosaehrznal ( horizontally are photons dent where igeoperator single states. entangled a purification mixed to entanglement from infor- and it quantum multiple-qubits apply encoding in as to mation such possible problems, therefore of range is larger wide It po- into one-qubit integrated filter. a be like larization should can just it filter networks, sepa- information device, our in quantum the output Because of pho- photon ports two both side. rate correct where the same cases produce where the reliably the 21], on eliminate [14, exit to splitter tons necessary beam is polarizing it post- a be- output difference by on realized significant function selection a the is and filter this this that tween Note indicates in desired output. post-selection photons no the the ancilla requiring in operation, the de- filter of are successful that Detection a photons probes not. input or as two filtering states photons the qubit ancilla two whether two this tect achieves using filter by The effect 1C). even (Fig. through and it passed preserve photons can the individ- between it entanglement the photons, create to the sensitive not of is polarization filter possibili- ual this two As 1B). these between (Fig. ties coherence quantum decreas- the the without share ing polarization they vertical if or only horizontal pairs same photon transmits that device photons. ancillary re- the of also that use will the anticipate Hilbert required quire may dimensional interactions one qubit-qubit higher and necessary a multi-qubits to polarization of a extended space of be concept can The optical filter 20]. required 19, the induce [18, pho- to non-linearity ancillary detection using photon by and realized tons been also have states ber w htniptsae[5]: state input photon two eateto lcrcladEetoi Engineering, Electronic and Electrical of Department & y 1 h prto fti le a esmaie ya by summarized be can filter this of operation The a is [5] in proposed filter polarization two-photon The n hgk Takeuchi Shigeki and ui-ui neatos edemonstrate We interactions. qubit-qubit h 3, d plcto.Fleigo h ai of basis the on Filtering application. ide | HH ∗ ain oqatmtechnologies. quantum to cations hyhv h eie orltosof correlations desired the have they f ogrF Hofmann, F. Holger n od rso,B81B UK 1UB, BS8 Bristol, Road, and st,Spoo0001,Japan 060-0812, Sapporo rsity, tmifrainsineadtech- and science information ntum i S ˆ and = S 4 1 ˆ | V V ( | eciigteeeto narbitrary an on effect the describing HH i eoe h tt hr ohinci- both where state the denotes ih Japan ,2, 1, HH † | − | 4 H n etcly( vertically and ) V V ih V V | ) , V po- ) (1) 2

A schematic of the optical quantum circuit for the en- tanglement filter Fig. 2A shows that multi photon inter- ferences occur at each of the four beam splitters BS1 to BS4. The crucial ones are BS2 and BS3. The working principle of the quantum circuit is as follows: whenever one of the ancilla photons meets a single input photon at BS2 or BS3, two photon quantum interference between the reflection of both photons and the exchange of an- cilla and input photon reduces the probability of finding a single photon at D1 or D2 to zero (Fig. 2A, inset) [22]. Thus, a single horizontally polarized photon cannot pass the circuit, resulting in the elimination of the HV and VH components. However, the HH component| i | i | i FIG. 1: The function of a polarization filter and a entangle- can pass because the two input photons will both ar- ment filter. (A) Polarization filter pass only the certain polar- rive at BS2 or BS3 together (the two H photons undergo ization component of single photons. (B) Entanglement filter quantum interference at BS1). In that case, the negative pass a pair of photons only if they share the same horizontal corresponding to the exchange of or vertical polarization. (C) As the quantum coherence be- tween these two possibilities is preserved during the process, the ancilla photon and one of the input photons is higher the output state is entangled when two diagonally polarized than the positive probability amplitude for three photon photons are input. reflection, and the HH component is transmitted with a flipped phase. | i To realize the quantum optical circuit of Fig. 2A, we larized, respectively. This filters out the HV have combined two recent technological approaches in or- and VH components of the two photon state| with-i | i der to simplify and stabilize quantum optical circuits: out reducing the coherence between the remaining HH the displaced-Sagnac architecture [6] and partially po- and V V components. The negative sign between| thei | i larizing beam splitters (PPBSs) [7, 8, 9]. The PPBSs terms is a consequence of the phase difference between reflect vertically polarized photons perfectly while trans- two photon reflection and two photon transmission at mitting/reflecting horizontally polarized photons with the beam splitters. The factor of 1/4 is an expression of 50% probability. Replacing all the beam splitters (BS1 to the transmission probability of 1/16 for HH and V V . | i | i BS4) with such PPBSs, we can remove the four polariz- It should be noted however that the detection of the an- ing beamsplitters used to separate the two polarizations cilla photons indicates a successful transmission of the (Fig. 2B). The remaining optical path interferometer is two input photons, so that it is not necessary to detect realized as a displaced Sagnac interferometer (Fig. 2C). the output photons in order to know that they were suc- In this setup, all the four polarization modes of two input cessfully transmitted. photons pass through all the optical components inside To understand the unique quantum properties of this the interferometer so that the path-differences between filter, it is useful to consider the effects of the filtering those four polarization modes are robust against drifts process on two diagonally polarized photons. As diag- or vibrations of optical components. onal polarizations are coherent superpositions of hor- We used photons generated via type-I spontaneous izontal and vertical polarizations ( P 1 ( H + | i ≡ √2 | i parametric down-conversion [23]. The pump laser pulses V ); M 1 ( H V )), two diagonally polar- (82 MHz, @ 390 nm) pass through a beta-barium borate | i | i ≡ √2 | i − | i ized photons have a well defined coherence between (BBO) crystal twice to generate two pairs of photons. their HH and V V components that determines Two photons in one of the pairs are used as the signal whether| theiri diagonal| polarizationsi are parallel (+) [eg: photons, and the two photons in the other pair are used 1 P P / MM = 2 ( HH + V V ...)] or orthogonal as the ancillary photons. The visibility of the Hong-Ou- | i | i | i1 | i± ( )[eg: PM / MP = 2 ( HH V V ...)]. The op- Mandel dip [22] was 96 1% for two photons in the same − ˆ| i | i | i−| i± pair, and 85 5% for photons± from different pairs. To test erator S preserves the magnitude of this coherence, but ± flips the sign: Eq. 1. Therefore, the correlation between the performance of our quantum filter circuit, we used co- the diagonal polarizations is inverted; parallel inputs re- incidence measurements between the four threshold de- sulting in the superpositions of the orthogonal outputs tectors at D1, D2 and the two outputs, rather than using and vice versa. Moreover, the coherence between HH photon number discriminating detectors [24, 25, 26, 27] and V V indicates a correlation between the circular| i for D1 and D2, as we needed to analyze the polarization polarizations—opposite| i directions (+) or same direction state of the output to confirm correct operation. This po- ( ). Thus, filtering out the HV and VH components larization analysis was performed using pairs of half of− two diagonally polarized input| i photons| alsoi generates plates and quarter wave plates together with polarizing correlations between the circular polarizations of the pho- beam splitters. tons, indicating the generation of entanglement. First, we checked the essential operation of the filter 3

FIG. 2: Optical quantum circuit for the nondestructive entanglement filter. (A) The original circuit includes concatenated path-interferometers together with four quantum interferences. PBS: Polarizing Beam Splitter which reflects (transmits) vertical (horizontal) polarization component, BS: Beam Splitter whose reflectance/transmittance is 1/2 for horizontal/vertical polarization. (inset) two photon interference at BS2 and BS3. Blue (red) line indicates optical paths for vertically (horizontally) polarized components. (B) A semi-simplified circuit using partially polarizing beam splitters (PPBS). The reflectance of PPBS A for horizontally and vertically polarized photons are 1/2 and 1, respectively. The transmittance of PPBS B for horizontally and vertically polarized photons are 1/2 and 1, respectively. (C) The final simplified system used in our experiment. Now the system includes one super-stable optical path-interferometer realized using displaced Sagnac architecture. circuit: we prepared input signal photons in the four com- polarizations (Y-basis). The predicted correlations be- binations of horizontal (H) and vertical (V) polarizations tween the circular polarizations of the output are ob- (which we call the Z-basis states) and observed how those served with a fidelity of FX Y =0.68. To complete the input states are filtered by the circuit (Fig. 3). It is clear test of entanglement generation,→ we also detected the di- from the experimental data that the photon pairs are agonal polarizations of the output to test whether the transmitted through the filter when the two input pho- polarization correlation is correctly flipped by the filter. tons share the same polarization (HH or VV), and most The fidelity of this filter operation is FX X = 0.60. of the pairs are filtered out when the two input photons Taken together, the three experimental tests→ are suffi- have different polarization (HV or VH). The fidelity of cient for the evaluation of the quantum filter operation. this process can be defined as the ratio of correctly trans- Because almost all the errors conserve horizontal/vertical mitted photon pairs to the total number of transmitted polarization (see Appendix 1), the probability of the cor- photon pairs. For the filter’s operation on horizontal and rect operation Sˆ is given by the process fidelity vertical polarizations, the fidelity is FZ Z =0.80. → Fp = (Fz >z + Fx >y + Fx >x 1)/2=0.54. (2) Next, it is necessary to test the preservation of quan- − − − − tum coherence between the HH and the V V com- The entanglement capability of the filter can be evaluated | i | i ponents transmitted by the filter. As explained above, as C = 2Fp 1=0.08, and so our experimental results this can be done by using diagonally polarized photons provide clear− evidence of the entangling capability of the in the input. We prepared pairs of photons in all combi- quantum filter operation. nations of diagonal linear polarizations P and M (the The entanglement filter will be a key element in the X-basis states) as signal inputs. The input| i photons| i are control of multi-photon quantum states, with a wide then in an equal superposition of the four different hor- range of applications in entanglement based quantum izontal/vertical polarization states HH , HV , VH , communication and quantum information processing. and V V (with different signs of the| coherencesi | i for| eachi For the present tests of the performance of the quantum of the| fouri different inputs). The filter should transmit filter circuit, we used threshold detectors to monitor the only the HH and the VV components while preserving output state. For applications where the output state the quantum coherence between them. As a result, the cannot be monitored, high-efficiency number-resolving ideal output states are entangled states with maximal photon detectors [24, 25, 26, 27] could be used at D1 correlations in both the circular and the diagonal polar- and D2 to generate the heralding signals. Such a circuit izations. To test this entanglement generation, we first could be used for on-demand generation of entangled detected the output photons in the right-circular R photons, non-destructive entanglement purification, and ( H + i V )/√2 and left-circular L ( H i V |)/i√ ≡2 could be implemented using an integrated waveguide | i | i | i≡ | i− | i 4

ture 421, 343 (2003). [16] B. Misra, E. C. G. Sudarshan, J. Math. Phys. 18, 756 (1977). [17] P. G. Kwiat, et al., Phys. Rev. Lett. 83, 4725 (1999). [18] K. Sanaka, Phys. Rev. A 71, 021801 (2005). [19] K. Sanaka, K. J. Resch, A. Zeilinger, Phys. Rev. Lett. 96, 083601 (2006). [20] K. J. Resch, et al., Phys. Rev. Lett. 98, 203602 (2007). [21] T. B. Pittman, B. C. Jacobs, J. D. Franson, Phys. Rev. A 64, 062311 (2001). [22] C. K. Hong, Z. Y. Ou, L. Mandel, Phys. Rev. Lett. 59, 2044 (1987). [23] D. C. Burnham, D. L. Weinberg, Phys. Rev. Lett. 25, 84 FIG. 3: Experimental results. Input signal photons are pre- (1970). pared in horizontal (H) or vertical (V) polarization and are [24] S. Takeuchi, J. Kim, Y. Yamamoto, H. H. Hogue, Appl. measured on H/V basis. (A)Theoretically predicted proba- Phys. Lett. 74, 1063 (1999). bilities (B) four fold coincidence count rates [counts/800s] of [25] J. Kim, S. Takeuchi, Y. Yamamoto, H. H. Hogue, Appl. the four detectors D1 to D4 are shown. Note that the events Phys. Lett. 74, 902 (1999). where two pairs of photons are simultaneously incident to the [26] E. J. Gansen, et al., Nature Photon. 1, 585 (2007). ancillary inputs and no photons are incident to the signal in- [27] B. E. Kardynal, Z. L. Yuan, A. J. Shields, Nature Photon. puts are subtracted, as confirmed by a reference experiment 2, 425 (2008). without input photons. ( 6 coincidence counts / 800s for each [28] A. Politi, M. J. Cryan, J. G. Rarity, S. Yu, J. L. O’Brien, of four cases with HH outputs. ) Science 320, 646 (2008). architecture [28]. This work was supported by the Japan Science and Technology Agency (JST), Ministry of Internal Affairs and Communication (MIC), Japan Society for the Promotion of Science (JSPS), 21st Century COE Program, Special Coordination Funds for Promoting Science and Technology.

∗ These authors contributed equally † Electronic address: [email protected] [1] N. Gisin, G. Ribordy, W. Tittel, H. Zbinden, Rev. Mod. Phys. 74, 145 (2002). [2] M. A. Nielsen, I. L. Chuang, Quantum Computation and Quantum Information (Cambridge University Press, 2000). [3] V. Giovannetti, S. Lloyd, L. Maccone, Science 306, 1330 (2004). [4] A. N. Boto, et al., Phys. Rev. Lett. 85, 2733 (2000). [5] H. F. Hofmann, S. Takeuchi, Phys. Rev. Lett. 88, 147901 (2002). [6] T. Nagata, R. Okamoto, J. L. O’Brien, K. Sasaki, S. Takeuchi, Science 316, 726 (2007). [7] N. K. Langford, et al., Phys. Rev. Lett. 95, 210504 (2005). [8] N. Kiesel, C. Schmid, U. Weber, R. Ursin, H. Weinfurter, Phys. Rev. Lett. 95, 210505 (2005). [9] R. Okamoto, H. F. Hofmann, S. Takeuchi, K. Sasaki, Phys. Rev. Lett. 95, 210506 (2005). [10] N. Gisin, R. Thew, Nature Photon. 1, 165 (2007). [11] B. L. Higgins, D. W. Berry, S. D. Bartlett, H. M. Wise- man, G. J. Pryde, Nature 450, 393 (2007). [12] M. D’Angelo, M. V. Chekhova, Y. Shih, Phys. Rev. Lett. 87, 013602 (2001). [13] J. L. O’Brien, Science 318, 1567 (2007). [14] J.-W. Pan, C. Simon, C.¨ Brukner, A. Zeilinger, Nature 410, 1067 (2001). [15] T. Yamamoto, M. Koashi, S. K. Ozdemir,¨ N. Imoto, Na- 5

Appendix 1 operation and Sˆxy and Sˆxx give false results. Therefore, the fidelities Fi j can be given by the sums of the prob- → Derivation of the process fidelity. The errors due to ability Fp = χ0,0 for the correct operation Sˆ0 and the the change of input horizontal/vertical polarization (i.e. probabilities ηij = χnn for the errors Sˆn as follows. HH HV ) are negligibly small in our experiment | i → | i shown in Fig. 3B ( Z Z measurement ). Based on this Fz z = Fp + ηzz fact, we can assume that→ the input horizontal/vertical po- → Fx y = Fp + ηxy larization is preserved in our filter, and thus our filter op- → Fx x = Fp + ηxx (5) eration can be well described in terms of superpositions → of HH HH , HV HV , VH VH and V V V V . Note that these relations between the diagonal elements Under| thisih assumption,| | ih we| can| estimateih | the| processih fi-| of the process matrix and the experimentally observed delity of our entanglement filter as follows. fidelities can also be derived from eq. (4) using the formal It is convenient to express the errors in terms of the definition of the experimental fidelities. In this case the superpositions fidelities are determined by the sums over the correct ˆ outcomes (j)l in E( (i)k (i)k ), averaged over all inputs S0 = √2( HH HH V V V V ), | i | ih | | ih | − | ih | (i)k , Sˆzz = √2( HH HH + V V V V ), | i | ih | | ih | ˆ √ Sxy = 2( HV HV + VH VH ), Fi j = X (j)l E( (i)k (i)k ) (j)l /4) (6) | ih | | ih | → h | | ih | | i l,k Sˆxx = √2( HV HV VH VH ). (3) | ih | − | ih | = χ ( (j) Sˆ (i) (i) Sˆ (j) /4)(7). ˆ X nm X l n k k m l Here, S0 represents the intended operation of the entan- n,m h | | ih | | i ˆ l,k glement filter, Szz represents the operation with a phase Here k,l 1, 2, 3, 4 , and (i)k denotes the k th state of ˆ ˆ ∈{ } flip error between HH and V V , and Sxy, Sxx repre- the i basis states. For example, (i)1 = HH , (i)2 = | i | i | i | i | i sent the leakage (or transmission error) of the HV and HV , (i)3 = VH , (i)4 = V V for i = z. The sums VH components, either with or without a| phasei flip | i | i | i | i | i | i over initial states k and final states l are one for n = between these components. The operation of our filter m = 0 and for n = m = ij and zero in all other cases, can then be written as confirming the results in eq.(5). Since the diagonal elements of the process matrix E(ρ )= χ Sˆ ρ Sˆ (4) in X nm n in m correspond to the probabilities of the orthogonal ba- n,m sis operations, their sum is normalized to one, so that χ = F + η + η + η = 1. It follows that the where n,m 0, zz, xy, xx and χnm are the process Pn nn p zz xy xx matrix elements∈ { of the noisy} quantum process. sum of all three experimentally determined fidelities is Fz z + Fz y + Fx x =2Fp + 1. Therefore, the process Each of our experimentally observed truth table op- → → → erations i j is correctly performed by Sˆ0 and one fidelity of our circuit is given by → other operation Sˆn, as indicated by the index n = ij for i, j x,y,z . For example, in case of the z z Fp = (Fz z + Fx y + Fx x 1)/2=0.54. (8) ∈ { } → → → → − truth table, operation Sˆ0 and Sˆzz result in the correct